Remote Sensing-based Yield Curves (RSYC) for Canada’s Forested Ecozones
Carbon accounting
National Forest Carbon Monitoring, Accounting and Reporting System (NFCMARS) is designed to:
estimate past changes in forest carbon stocks, such as from 1990 to the present (monitoring)
predict changes in carbon stocks, based on scenarios of future disturbance rates and management actions, in the next 2 to 3 decades (projection)
Existing models - limitations
spatial extent (single jurisdiction, Canada-wide)
developed for the dominant (most common) species only
developed for managed forests
often based on a limited number of samples
Data
- Landsat-derived AGB time series (1984-2021) (Matasci et al. 2018)
- Landsat-derived age (Maltman et al. 2023)
- Landsat-derived dominant species (Hermosilla et al. 2022)
Tiling
Approach
- Build a chronosequence of AGB by combining multiple pixel-level AGB time series
- Develop a filtering approach to screen Landsat-derived AGB
- Fit non-linear mixed effects models (NLME)
Approach
Modeling
Model form after (Fortin and Lavoie 2022):
\[ AGB = \beta_1 e^{-\beta_4 age} (1 - e^{-\beta_2 age})^{\beta_3} \]
NLME:
- nested tile/pixel random effect
- random parameter: \(\beta_1\)
Results
National AGB yield curves (N=1892)
Multi-species:
- Generic (all species) – 389
- Coniferous – 355
- Broadleaf – 184
Species-specific:
- 27 species-specific models (some for a single tile only, up to 308 for black spruce)
Model evaluation
Yield curves represent the average AGB in the tile
Evaluation – compare mean AGBRSYC with AGBREF by 20 year age classes
Model evaluation
MAGPlot data:
Over 50k plots, harmonized dataset
Independent
Mix of different sampling designs
Raking ratio: weights (per plot) to adjust for under- or over-representation of different forest strata (e.g. plots located in more productive stands)
